OpenCV 4 with Python Blueprints - Second Edition

Learning about PCA

PCA is a dimensionality-reduction technique that is helpful whenever we are dealing with high-dimensional data. In a sense, you can think of an image as a point in a high-dimensional space. If we flatten a 2D image of height m and width n (by concatenating either all rows or all columns), we get a (feature) vector of length m x n. The value of the i^th element in this vector is the grayscale value of the i^th pixel in the image.

To describe every possible 2D grayscale image with these exact dimensions, we will need an m x n-dimensional vector space that contains 256^{m x n} vectors. Wow!

An interesting question that comes to mind when considering these numbers is—Could there be a smaller, more compact vector space (using less-than ...